UWF EVR 6930 - Monitoring land Cover Changes

Unformatted text preview:

int. j. remote sensing, 1999, vol. 20, no. 1, 139 ± 152Monitoring land-cover changes: a com parison of changedetection techniques*J.-F. MASLaboratory of Remote Sensing and Geographic Information Systems,Centre of Ecology, Fisheries and Oceanography of the Gulf ofMexico (EPOMEX), University of Campeche, AP 520 24030 Campeche,CAMP, Mexico; e-mail: [email protected](Received 20 May 1997; in ® nal form 20 May 1998)Abstract.Six change detection procedures were tested using Landsat Multi-Spectral Scanner (MSS) images for detecting areas of changes in the region ofthe TeÂrminos Lagoon, a coastal zone of the State of Campeche, Mexico. Thechange detection techniques considered were image dierencing, vegetative indexdierencing, selective principal components analysis (SPCA), direct multi-dateunsupervised classi® cation, post-classi® cation change dierencing and a combina-tion of image enhancement and post-classi® cation comparison. The accuracy ofthe results obtained by each technique was evaluated by comparison with aerialphotographs through Kappa coecient calculation. Post-classi® cation compar-ison was found to be the most accurate procedure and presented the advantageof indicating the nature of the changes. Poor performances obtained by imageenhancement procedures were attributed to the spectral variation due to dier-ences in soil moisture and in vegetation phenology between both scenes. Methodsbased on classi® cation were found to be less sensitive at these spectral variationsand more robust when dealing with data captured at dierent times of the year.1. IntroductionSeveral regions around the word are currently undergoing rapid, wide-rangingchanges in land cover. Much of this activity is centred in the tropics in such countriesas Brazil, Columbia, Indonesia, Mexico, the Ivory Coast, Venezuela and Zaire (FAO1995). These changes in land cover, in particular tropical forest clearing, haveattracted attention because of the potential eects on erosion, increased run-oand¯ ooding, increasing CO2concentration, climatological changes and biodiversity loss(Myers 1988, Fontan 1994). Remote sensing provides a viable source of data fromwhich updated land-cover information can be extracted eciently and cheaply inorder to inventory and monitor these changes eectively. Thus change detection hasbecome a major application of remotely sensed data because of repetitive coverageat short intervals and consistent image quality.The basic premise in using remote sensing data for change detection is that*Presented at the Fourth International Conference on Remote Sensing for Marine andCoastal Environments, Orlando, Florida, 17± 19 March 1997.International Journal of Remote SensingISSN 0143-1161 print/ ISSN 1366-5901 onlineÑ1999 Taylor & Francis Ltdhttp://www.tandf.co.uk/JNLS/ res.htmhttp://www.taylorandfrancis.com/JNLS/ res.htmJ.-F. Mas140changes in land cover result in changes in radiance values and changes in radiancedue to land cover change are large with respect to radiance changes caused by othersfactors such as dierences in atmospheric conditions, dierences in soil moisture anddierences in sun angles. Vegetation diversity and interspersion of land cover is highin the humid tropics, and spectral re¯ ectance characteristi cs of mixed vegetation areoften not distinct, causing problems in digital classi® cations (Royet al.1991, Saderet al.1991). For example, workers have reported spectral confusion between undis-turbed and disturbed forests (Franklin 1993) and between successional forest classesand pasture containing trees (Saderet al.1990). Similar confusion is also expectedwhen trying to discriminate natural grassland (savannah) from pasture lands.Therefore, change between land covers which present similar spectral signatures isdicult to detect.The impact of sun angle dierences and vegetation phenology dierences maybe partially reduced by selecting data belonging to the same time of the year (Singh1989). However, it may be extremely dicult to obtain multi-date data of the sametime of the year, particularly in tropical regions where cloud cover is common. Forexample, one of the objectives of the North American Landscape Characterization(NALC) Project was to produce `triplicates’ consisting of Landsat Multi-SpectralScanner (MSS) images for the years 1973, 1986 and 1992 (plus or minus one year)for the USA and Mexico. In Mexico, because of the low number of good qualityscenes, images of the triplicate often do not belong to the same time of the year.Moreover, in many cases it had even been necessary to get images dated outside the3-year intervals previously de® ned and to mosaic various images to obtain a cloud-free scene. Table 1 shows that dierences between the dates of the NALC triplicateimages for southern Mexico (path 18 to 25) are, on average, higher than 2 months.The problem of availability of cloud-free images in tropical regions is very commonand has been reported by many authors (Jha and Unni 1982, Ducros-Gambart andGastellu-Etchegorry 1984, Nelson and Holben 1986, Pilonet al.1988, Alwasheand Bokhari 1993) . Consequently, the identi® cation of a robust change detection-methodology is essential for dealing with multi-date data in these regions.An analysis of the literature reviewed indicates that (1) there are very few studiesconcerned with comparative evaluation of change detection techniques, (2) themajority of these comparative studies have not supported their conclusion by quantit-ative analysis of the results (Singh 1989) and (3) many studies have been carried outwith images of the same time of the year. The purpose of this study is to comparethe relative eectiveness of dierent techniques in detecting land cover changes in atropical coastal zone using images captured at dierent times of the year.Table 1. Dierences in the dates of NALC triplicate images of south-eastern Mexico. Valuesare expressed in number of days of dierence.Decades1970± 1980 1970± 1990 1980± 1990Mean 68.2 81.3 59.2Maximum 182 177 154Minimum 1 2 3SD 51.9 50.6 48.7SD, standard deviation.Monitoring lan d-cover changes1412. Study areaThe area of study covers a part of the Te rminos Lagoon region, State ofCampeche, Mexico. It is located in the south-east of Mexico, between 18ß00¾and18ß55¾N latitude and 90ß55¾and 92ß06¾W longitude (® gure 1). Land use withinthis area is divided principally among mangrove, evergreen tropical


View Full Document

UWF EVR 6930 - Monitoring land Cover Changes

Documents in this Course
Load more
Download Monitoring land Cover Changes
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Monitoring land Cover Changes and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Monitoring land Cover Changes 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?